WIXLIB

ERIC NOTEBOOK SERIESSecond EditionCross-sectional StudiesSecond Edition Authors:Lorraine K. Alexander, DrPHBrettania Lopes, MPHKristen Ricchetti-Masterson, MSPHKarin B. Yeatts, PhD, MSLike cohort studies, cross-sectionalstudies conceptually begin with apopulation base. But unlike cohortstudies, in cross-sectional studieswe do not follow individuals overtime. Instead, we only look at theprevalence of disease and/orexposure at one moment in time.These studies take a "snapshot" ofthe proportion of individuals in thepopulation that are, for example,diseased and non-diseased at onepoint in time. Other healthoutcomes besides diseases mayalso be studied. Cross-sectionalstudies also differ from cohortstudies in the populations that arestudied. Cohort studies begin byselecting a population of personswho are at risk of for a specificdisease or health outcome; crosssectional studies begin by selectinga sample population and thenobtaining data to classify allindividuals in the sample as eitherhaving or not having the tion-atriskEntire population (or asample)CommonMeasuresRisks andRatesPrevalenceWays to use cross-sectional studiesCross-sectional studies are used bothdescriptively and analytically.Descriptive cross-sectional studiessimply characterize the prevalence ofa health outcome in a specifiedpopulation. Prevalence can beassessed at either one point in time(point prevalence) or over a definedperiod of time ( period prevalence).Period prevalence is required when ittakes time to accumulate sufficientinformation on a disease in apopulation, e.g. what proportion ofpersons served by a public healthclinic over a year have hypertension.These prevalence measures arecommonly used in public health;often the point or period aspect is notspecified.In analytical cross-sectional studies,data on the prevalence of bothexposure and a health outcome areobtained for the purpose of comparinghealth outcome differences betweenexposed and unexposed.Analytical studies attempt to describethe prevalence of, for example,disease or non-disease by firstbeginning with a population base.These studies differ from solelydescriptive cross-sectional studies inthat they compare the proportion ofERIC at the UNC CH Department of Epidemiology Medical Center

ERIC NOTEBOOKPA G E 2exposed persons who are diseased (a/(a b)) with theproportion of non-exposed persons who are diseased (c/(c d)).Calculating prevalenceThe prevalence of a health outcome is simply theproportion of individuals with the health outcome in apopulation.Prevalence cases / total populationFor the following example, two different sub-measures ofprevalence can be calculated: the prevalence of coronaryheart disease (CHD) among the exposed (people who arenot active) and the prevalence of CHD among theunexposed.time for the case-control study. The PR is similar to a RRwhen the outcome occurs over a short period of time. Forexample, one would calculate a prevalence ratio for anacute outbreak of tuberculosis in a prison population.This is in contrast to calculating the overall prevalence ofpositive tuberculin skin tests among the prisoners.The prevalence ratio can also be calculated from theinformation on CHD and physical activity. It is preferableto calculate the prevalence odds ratio when the period forbeing at risk of developing the outcome extends over aconsiderable time (months to years) as it does in thisexample:PR (a/N1) / (c/N0)PR (50/250) / (50/750) 3.0PresentCHD50aAbsent CHDTotalIn this case, a prevalence ratio of 3.0 can be interpretedto mean that the proportion of people with CHD is 3-foldgreater if a person is not physically active.b200250POR vs. PRActive50d700750Total1009001000For chronic disease studies or studies of long-lasting riskfactors, POR is the preferred measure of association incross-sectional studies. For acute disease studies, PR isthe preferred measure of association. If the prevalence ofdisease is low, i.e. 10% or less in exposed and unexposedpopulations, POR PR. Since cross-sectional studies areparticularly useful for investigating chronic diseases (e.g.prevalence of AIDS) where the onset of disease is difficultto determine, or for studying long lasting risk factors (suchas smoking, hypertension, and high fat diets), theprevalence odds ratio will generally be the preferredmeasure of association.Example:Not activecP1 a/a b 50/250 20.0% prevalence ofCHD among people who are not active.P0 c/c d 50/750 6.7% prevalence ofCHD among people who are active.The prevalence odds ratioThe prevalence odds ratio (POR) is calculated in the samemanner as the odds ratio.POR ad / bcLimitations of cross-sectional studies to evaluate riskRecall that, under steady conditions, the prevalence ofdisease is influenced both by incidence and duration ofdisease (or survival with disease).The prevalence ratioThe prevalence ratio (PR) is analogous to the risk ratio (RR)of cohort studies. The denominators for both ratios arefixed populations – fixed at the start of the study in thecase of a cohort study, and fixed at the point or period ofPrevalence Rate x Average Duration of DiseaseERIC at the UNC CH Department of Epidemiology Medical Center

ERIC NOTEBOOKPersons who survive longer with a disease will have ahigher probability of being counted in the numerator of aprevalence proportion. Short-term survivors will be lesslikely to be counted as a case. Incidence is influenced onlyby exposure, whereas prevalence is influenced both byexposure and duration of disease.If exposure influences survival time, then the POR or PRwill not provide a valid estimate of the risk ratio or rateratio. Thus, the interpretation of the POR or PR is subjectto survival bias.Even if incidence remains constant, either animprovement in disease treatment (that results in highercure rates) or increased lethality (resulting in a higher casefatality rate) will result in decreased prevalence. Thedisease itself or the threat of developing the disease maycause outmigration of cases from an environmentperceived as causing disease, e.g. workers affected bytoxic exposures in a plant may quit, while more resistantworkers will stay. This selective migration can biasmeasures of prevalence.Other problems with interpretation of cross-sectionalstudiesCross-sectional studies as well as case-control studies areaffected by the antecedent-consequent bias, similar to thechicken and egg question (i.e. “which came first?”). Thisbias occurs when it cannot be determined that exposurepreceded disease, since both are ascertained at the sametime (unlike cohort studies or clinical trials). Antecedentconsequent bias does not affect cohort studies becausesubjects in cohort studies are selected for study becausethey are disease-free. Exposure is actually observed toprecede disease only in a cohort design, includingrandomized trials.Uses of cross-sectional studiesPA G E 3Descriptive cross-sectional studies are widely used toestimate the occurrence of risk factors in segments of thepopulation characterized by age, sex, race orsocioeconomic status (SES). National examples of crosssectional studies of great importance are the decennialcensus and the National Health and Nutrition Surveys(NHANES). Opinion polls and political polls are basicallycross-sectional studies. Surveillance of changes insmoking habits or of other behavioral risk factors aresequential cross-sectional studies. The US NationalHealth and Nutrition Examination Survey (NHANES) is onesuch example Similarly, surveillance of long lastingdiseases such as AIDS is cross-sectional. Descriptivecross-sectional studies are useful for planning oradministering preventive or health care services,surveillance programs, and surveys and polls.Descriptive/analytical cross-sectional studies are usefulfor establishing preliminary evidence for a causalrelationship. These studies are also useful for examiningthe association between exposure and disease onset forchronic diseases where researchers lack information ontime of onset. Examples might include diet and arthritis,smoking and chronic bronchitis, and asthma andexposure to air pollution. Interpretation requires cautionregarding potential association of duration of disease withexposure status (survival bias).Survival bias may be minimized if information can beobtained on exposures that clearly preceded the firstsymptoms of a chronic disease such as arthritis, diabetes,or chronic bronchitis. This depends on access to medicalrecords before the onset of a chronic disease. In addition,it may be necessary to have historical records on anindividual’s exposure status prior to these first medicalvisits, e.g. where the person lived or where the person wasemployed.Descriptive studies are an important method to evaluatethe proportion of a population with disease or with riskfactors for disease, such as the prevalence of asthma inchildren or the prevalence of elevated blood lead intoddlers.ERIC at the UNC CH Department of Epidemiology Medical Center